Fix step used to log SAC summary

[krishpop]

fixes hill-a#1007

Description

Changed call to writer.add_summary to use self.num_timesteps in the case that num_timesteps is not reset after multiple calls to model.learn(..., reset_num_timesteps=False).

Motivation and Context

• I have raised an issue to propose this change (required for new features and bug fixes)
  Issues Tensorboard Logging Issue after multiple consecutive calls to self.learn(..., reset_num_timesteps=False) hill-a/stable-baselines#1007 shows an example where tensorboard plotted line is discontinuous due to incorrect step counter

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)
• Documentation (update in the documentation)

Checklist:

• I've read the CONTRIBUTION guide (required)
• I have updated the changelog accordingly (required).
• My change requires a change to the documentation.
• I have updated the tests accordingly (required for a bug fix or a new feature).
• I have updated the documentation accordingly.
• I have ensured pytest and pytype both pass (by running make pytest and make type).